# Angze Plot Style > Apply Angze's evidence-backed Matplotlib scientific plotting conventions. Use when creating, restyling, reviewing, or exporting Python scientific figures for Angze, including manuscript/final figures, compact diagnostics, PDI compound comparisons, and plot-style decisions. - Skill: `angzeli/angze-plot-style` (Agent Skill, multi-file: 10 files) - Install (CLI): `npx skillmds@latest add angzeli/angze-plot-style` - Raw SKILL.md: https://api.skillmd.com/api/skills/angzeli/angze-plot-style/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: angzeli (https://skillmd.com/u/angzeli) - Updated: 2026-09-22 - Page: https://skillmd.com/skills/angzeli/angze-plot-style --- # Angze Plot Style Generate ordinary, self-contained Matplotlib code in Angze's established plotting language. Treat this skill as the runtime specification: normal output must not import, copy, install, or add a path for the bundled Python helper. ## Usage model 1. Identify the figure context before styling: - `base`: stable visual DNA for ordinary scientific plots. - `manuscript`: base DNA plus title-free, selective-legend, paired-export policy. - `diagnostic`: compact typography for working figures. 2. Write self-contained Matplotlib. For several related figures in one script, define a small local `style_axes(ax)` helper rather than depending on a personal external module. 3. Prefer one standalone `(8, 6)` scientific figure per logical result rather than combining unrelated plots into one giant panel. Use panels when the scientific comparison genuinely requires them. ## Base contract Use `figsize=(8, 6)`, a white ground, black foreground, Arial-first sans serif, and STIX Sans math. Apply at least these rcParams; do not set a global `font.size` or force `mathtext.default`: ```python { "figure.facecolor": "white", "axes.facecolor": "white", "savefig.facecolor": "white", "legend.facecolor": "white", "legend.edgecolor": "black", "font.family": "sans-serif", "font.sans-serif": ["Arial", "Helvetica", "Liberation Sans", "DejaVu Sans"], "mathtext.fontset": "stixsans", "text.color": "black", "axes.edgecolor": "black", "axes.labelcolor": "black", "axes.labelsize": 22, "axes.labelweight": "bold", "axes.titlesize": 18, "axes.titleweight": "bold", "axes.linewidth": 1.8, "axes.grid": False, "axes.spines.left": True, "axes.spines.right": True, "axes.spines.bottom": True, "axes.spines.top": True, "xtick.labelsize": 14, "ytick.labelsize": 14, "xtick.color": "black", "ytick.color": "black", "xtick.direction": "in", "ytick.direction": "in", "xtick.major.width": 1.8, "ytick.major.width": 1.8, "xtick.major.size": 4, "ytick.major.size": 4, "xtick.bottom": True, "ytick.left": True, "xtick.top": False, "ytick.right": False, "xtick.minor.visible": False, "ytick.minor.visible": False, "lines.linewidth": 2.0, "legend.fontsize": 10, "legend.frameon": True, "legend.framealpha": 1.0, "savefig.dpi": 600, "savefig.bbox": "tight", "savefig.edgecolor": "white", "savefig.transparent": False, } ``` ## Colour contract An explicit current user colour instruction always overrides this skill. Keep the exact hex, RGB, Matplotlib/named colour, identity mapping, palette, or order the user supplies; do not replace, correct, or semantically remap it. If only some colours are specified, preserve them exactly and apply this policy only to the remaining identities. For colour decisions, use this priority from highest to lowest: 1. explicit current user instruction; 2. an existing deliberate project/task semantic mapping; 3. the built-in exact semantic registry; 4. a mapping already established in the current figure set or workflow; 5. first-time assignment for new unrelated categorical identities; 6. the neutral/control role where scientifically appropriate. For new unrelated categorical identities, assign the first available colours in this exact order; do not select or reorder a local subset: ```python DEFAULT_COLOUR_CYCLE = [ "#0072B2", # blue "#D55E00", # orange "#7A5195", # purple "#009E73", # teal "#C23B70", # berry "#7A8F00", # olive ] NEUTRAL_COLOUR = "#4D4D4D" ``` Establish identity-to-colour mappings once per project, figure set, analysis, or related workflow and reuse them. A later subset retains the colours of its identities instead of being reassigned from the start of the cycle. Preserve a deliberate repository mapping unless the user requests restyling or replacement. Use `#4D4D4D` for a control, reference, literature benchmark, baseline, or neutral comparison when a categorical hue would imply an unnecessary identity; do not consume a cycle colour or treat neutral as a seventh category. Keep axes, text, spines, and suitable reference lines black as specified elsewhere. The exact semantic registry is identity-based, not plot-order-based: ```python PDI_COLOURS = { "PDI-Me-COOH": "#D55E00", "PDI-H-COOH": "#0072B2", "PDI-OMe-COOH": "#7A5195", } ``` For ordered values of one underlying identity—such as rate, temperature, time, loading, concentration, or potential—use a light-to-dark same-hue family from that identity's base colour instead of consuming unrelated categorical hues. Change the colour values themselves, not alpha. Use the frozen PDI rate shades for their exact contexts; derive other families locally and do not promote them to the canonical skill without later explicit approval. If more than six unrelated identities remain unmapped, do not invent, recycle, or import colours automatically. First consider scientifically meaningful line styles, marker shapes, direct labels, grouping, faceting, or separate figures; add hues only when the task requires them or the user supplies them. Do not use colour decoratively or change an identity's colour between related panels. Add redundant encoding when an important distinction would otherwise rely on colour alone, without inventing arbitrary marker variation. For every ordinary axes: - keep all four spines visible, black, and 1.8 pt wide; - disable both major and minor grids, then call `ax.minorticks_off()`; - use inward major ticks 1.8 pt wide and 4 pt long, with bottom/left ticks on and top/right ticks off while retaining their spines; - make every tick label 14 pt bold; - use 22 pt bold axis labels and 10 pt bold annotations; - draw data lines at 2.0 pt; - use the general marker `"o"`, 5.5 pt, filled with the series colour, with a black 0.8 pt edge; - leave scatter size, scatter alpha, and semantic marker shapes task-specific. Support an 18 pt bold title but do not display one by default. In reusable plotting code, provide a convenient commented line: ```python # ax.set_title("...", fontsize=18, fontweight="bold") ``` ## Legend, layout, and export Add a legend only when it communicates scientific information. Make it 10 pt bold with an opaque white frame and black edge, keep it inside the axes, and try `upper right`, then `upper center`, then `best` when collision requires it. Do not default to an outside legend. Use `fig.tight_layout()` for an ordinary figure. Use constrained layout for a genuinely dense or multipanel composition; do not impose universal subplot spacing. For final output, save only `.png` and `.pdf` by default. Save the PNG at 600 dpi and both with `bbox_inches="tight"`, `facecolor="white"`, and `transparent=False`. Do not emit SVG, EPS, TIFF, JPEG, or another format unless the user explicitly requests it. ## Context and optional resources Use quantity or descriptor plus units in parentheses; keep axis limits, scientific notation, decimal precision, significant figures, physical padding, annotation placement, scatter geometry, and reference-line width specific to the science. Use the compact diagnostic override only for working figures: `(5.8, 4.4)`, 13 pt bold labels, 10 pt bold ticks, 14 pt bold titles, and 9 pt bold legends and annotations. Retain the base axes geometry. Consult [references/style-profile.md](references/style-profile.md) for expanded detail and a self-contained example. Consult [references/colour-profile.md](references/colour-profile.md) for frozen PDI families, deterministic generic family generation, and worked colour-priority examples. Consult [references/exceptions.md](references/exceptions.md) only when spectra, dense panels, diagnostics, or another special context may justify an override. Consult [references/style-evidence.md](references/style-evidence.md) for provenance or future conflict review; explicit user decisions there override later mining until the user changes them. Treat [assets/angze_plot_style.py](assets/angze_plot_style.py) only as an optional reference implementation, executable specification, and validation fixture. Reuse it only when the user explicitly asks, a repository already vendors it deliberately, or the task is maintaining this skill.